I spent, golly, almost the last 3 weeks doing AGU online. Which was... fine, overall.
I mean, there's no way to do a 25,000-person conference virtually and have it be anywhere close to as good as the normal in-person version. But given that, it was not bad.
Posters were done in this "e-lighting" format where there are up to 6 boxes in a standard configuration, and you click on each box to expand it, so presenters can put as much information as they want into each box. I haaaaaaaate this format in person (presented on touch-screens), because it's so slow and clunky to click on each box, wait for it to expand, scroll through it, then back out and click on the next one, plus you can't scan quickly to see if there's anything you care about, and it drives me crazy. But then I discovered that viewing them on the web, there's a "print" view that just dumps everything onto a single long page, and that works just fine. (I think the lack of length and layout constraints in the e-poster format encourages people to make bad posters with poor information flow and organization, but honestly most people make bad posters in paper format, too, so it's not that much different.)
For talks, everyone giving an oral presentation had to pre-record their talk and upload it about a week before Thanksgiving. The videos were available to watch whenever you wanted, and then there was a scheduled zoom session for Q&A where the speakers each gave a brief 3-slide summary of their talk before answering any questions. Not all that different from a live presentation, and there's one thing I would take from the virtual conference back into the real world if I could: the ability to change the playback speed. Most presenters speak WAY too slowly, and being able to speed people up to 150% was fantastic. Plus you can pause and rewind if you need to.
With no time conflicts and the fact that I didn't have to spend any time going between things or waiting for things to start, I actually go to see a whole lot more AGU than normal. They have an electronic program where you can tick a box to add a presentation to your schedule, so I go through and add everything that looks interesting (based on keywords, people I know, and skimming through the session titles in the Atmospheric Science, Informatics, and Global Change secsion), and normally what happens is that I then have a jam-packed schedule for five days and I end up triaging away at least half of it on the day. But this time, I ended up skipping only a handful of things, and by the end of it, I had seen almost 260 talks and posters. (I exported my schedule and made a spreadsheet...) There were plenty of things that I spent about 30 seconds glancing over before writing "meh" in my notes, but most of what I saw was good.
The thing that I missed out on, of course, was people. I only went to one live poster session (for a friend from grad school, not even in my field), and I didn't run into anyone in the poster hall or the lobby or have any good post-talk conversations or meet anybody for lunch. So that was disappointing. But I don't think that was because it wasn't run well, I think that's just a limitation of the virtual format.
My only major complaint about the conference was the scheduling. Following some asinine reasoning about trying to be a "global scientific society", they tried to make the schedule friendly to Europe and Asia. Which means it was quite UNfriendly to all the North Americans making up the vast majority of the attendees. The oral Q&A sessions were all scheduled early in the morning or late in the evening. My talk was at 5:30 in the goddamn morning. I had to get up at 4-bloody-30-A.M. in order to be presentable and conscious on zoom from 5-6 (because I needed to be there for the whole thing, of course), speak for less than five minutes in the middle, and have time to get asked one (not particularly relevant) question near the end because the moderators didn't do a great job watching the clock. I was less than thrilled. (And I went straight back to bed afterward.)
But overall, it was pretty useful. I saw a lot of good stuff about machine learning, which has moved from "we think AI techniques might be useful for this kind of problem" to "we used deep learning for this process and got it to work well." So I've got my student assistant working on some ML stuff.
And this year I had the time to sit down afterwards and go through my notes to distill out the best stuff, so I'm hopeful that I'll put more of it into use. Not a bad note to end the work year on.
I still miss getting to go somewhere for it and see people and all that good stuff. Next year.